Blow flies are early colonizers of human remains, allowing their developmental rates to be used to estimate postmortem interval (PMI). Accurate species identification is essential for PMI estimation, yet traditional methods can be subjective. Recently, the use of cuticular hydrocarbons (CHCs) for more objective species identification has been demonstrated. These studies generally use multivariate statistical approaches to evaluate differences in CHC profiles among different species. Principal component analysis (PCA) is among the more common procedures used to evaluate these differences. However, PCA often requires data pretreatment steps to address inherent biological differences among specimens, which can result in the loss of meaningful biological trends. Further, few studies have developed the predictive classification models necessary for downstream PMI estimation. In this work, CHCs were extracted from adult females of seven blow fly species and analyzed by gas chromatography-mass spectrometry (GC-MS). Profiles were initially evaluated with PCA and non-metric multidimensional scaling (NMDS). From PCA, six of the seven species were distinguished, but distinction was influenced heavily by abundance differences due to inherent variability among specimens. From NMDS, tighter species clusters were observed without the need for data pretreatment. Classification models were then developed using partial least squares-discriminant analysis (PLS-DA) and random forest modeling. Both models demonstrated high classification success for cross validation (98.67% and 98.67%) and external validation (66.67% and 61.11%), although the random forest classifier had the advantage of requiring no data pretreatment. This work demonstrates successful species classification and highlights considerations for the statistical evaluation of CHC profiles.
Shirley et al. (Sun,) studied this question.